Imaging algorithms aid radiation therapy

At Masthead Imaging (Nanaimo, BC, Canada), the company`s Portal Imaging Processing Software (PIPS) package targets the problem of low-contrast portal images. Originally developed at the Manitoba Cancer Foundation (Manitoba, Canada) by Shlomo Shalev, the PIPS package can process, analyze, and display portal images obtained from digitized portal films or from an electronic portal imaging device.

Imaging algorithms aid radiation therapy

At Masthead Imaging (Nanaimo, BC, Canada), the company`s Portal Imaging Processing Software (PIPS) package targets the problem of low-contrast portal images. Originally developed at the Manitoba Cancer Foundation (Manitoba, Canada) by Shlomo Shalev, the PIPS package can process, analyze, and display portal images obtained from digitized portal films or from an electronic portal imaging device.

Not yet approved by the US Food and Drug Administration, the PIPS package includes a variety of contrast-limited adaptive histogram-equalization (CLAHE) techniques (see p. 88). These techniques are especially well suited to improving the contrast of portal images.

"In this technique," says Shalev," sharp field edges can be maintained by selective enhancement within the field boundaries." This enhancement is accomplished by first detecting the field edge in the portal image, and then processing only those regions of the image that lie inside the field edge. Noise is therefore reduced and the high spatial frequency content of the image is maintained by applying a combination of CLAHE, median filtration, and edge sharpening. "The technique, known as sequential processing, can be recorded into a user macro and reused at any time," adds Shalev.

Although contrast-enhancement techniques are relatively straightforward, a more complex imaging problem involves the measurement of patient set-up deviation between treatment sessions. To accomplish this, Martin Berger and his colleagues at the University Hospital of Zurich (Zurich, Switzerland) turned to the Generalized Hough Transform (GHT).

"To determine the patient displacement be tween a reference image and a treatment image," says Berger, "both field-edge and anatomy alignment have to be performed." Since computing a field-edge displacement is easier and more robust than performing anatomy alignment, it is performed first using a combination of a Canny edge detector and histogram equalizations.

After the field edges are computed, anatomy alignment is accomplished by searching the treatment image for significant structures taken from the reference image using small regions around the edges. "Because a single template is not deformed when using GHT, several small templates are manually defined to form a nonrigid master template that allows slight deformations between the single templates," says Berger.

More in Boards & Software